How to adapt materials for classroom diversity with AI

TLDR: Serving a diverse classroom with a single material is impossible, and producing three versions of every text by hand fits in nobody’s workday. Diffit adapts one text to several reading levels in minutes, always starting from your own source, ElevenLabs turns that same unit into a listenable version for students who struggle to read, and Gamma builds the visual supports. The professional rule: you adapt the path, never the learning goal.

This guide is for teachers with real classrooms: the group where a student reading two years below level sits next to a recent arrival to the language, a student who needs visual structure and one who is bored because it is all too easy. Adapting materials is among the clearest obligations of the job and the most impossible to meet by hand.

1. Map your group’s real needs

The classroom’s usual four profiles, each with the adaptation that serves it:

2. Generate the level versions with Diffit

Diffit is, among the classroom-tool catalog, the most specific one for this exact task: paste the unit’s text and it produces versions at the levels you request, always starting from your own source instead of inventing new content. Free plan useful enough to try, paid from $14.99 a month.

Correct use carries one nuance: check that the simplified version keeps the technical vocabulary that gets assessed. Simplifying syntax is fine. Replacing “photosynthesis” with “how plants eat” lowers the goal, and that is no longer adapting but cutting.

For adaptations it does not ship ready-made (versions for recent language arrivals with specific structures, say), what rescues the output is pasting your step 1 spec whole into the request: the tool’s documented weak spot is that it does not manage the whole class or replace the teacher’s review, and that review is exactly what decides whether the adapted version serves that profile.

3. Create the supports and alternative formats

Diversity is not just reading level: some students need to see the structure before reading, some process better through images and some get lost without a glossary. These supports, which by hand there was never time to produce, are now minutes each.

The unit’s advance glossary, each term explained in one sentence, gets generated from the text itself. The topic’s visual outline, which changes reading entirely for structure-dependent profiles, gets built with Gamma by pasting the content: its clean visual finish is exactly what a support needs. Free credits to start, paid from $10 a month.

The format most often forgotten is audio, and it is the one that levels most: the same unit made listenable for the student reading two years below, the one entering the language and the one with a diagnosed reading difficulty. ElevenLabs narrates the text in a natural voice from $5 a month, with a free plan to try it on one unit. The nuance that decides when to use it: where the assessment is comprehension of the content, listening is a legitimate path, and where the assessment is reading itself, audio accompanies but does not replace.

And for the student running ahead, extension stops being “read another text”: deepening activities generated on the same topic, with questions demanding connection rather than repetition.

4. Check that the adaptation does not lower the goal

The final review has one question: does every version allow reaching the same assessment criterion? If the adapted version no longer contains what you will assess, you have not adapted the material: you have quietly decided that student will not reach the goal, and that decision is not yours to make through the material.

The practical check: take the unit’s assessment criterion and verify each version contains what is needed to work on it.

Legitimate differences live in syntax, scaffolding and format. The content that gets assessed stays.

The adapted assessment for those same units gets built with the exams and quizzes guide, which can generate variants of the same exam with different scaffolding. The rest of the teaching cycle is in the lesson planning guide and the AI for education and training hub.

Common mistakes

Adapting by lowering the goal. Simplifying the path is differentiation. Emptying the assessable content is condemning in advance. The step 4 question gets asked every time.

Generating adaptations without a map. Producing generic “easy” versions without knowing what each profile needs is wasted work that happens to look like inclusion.

Labeling the versions visibly. Handing out the “easy text” in front of the whole group stigmatizes. Same design, same cover, different interior.

Bypassing counseling on significant adaptations. Tools produce materials in minutes, and that speed does not turn the teacher into a counseling team. Diagnosed plans govern.

Frequently asked questions

How long does adapting a full unit take?

With the map done, producing three levels of a text, a glossary and a visual outline takes under an hour, against the full afternoon per version it took by hand. Reviewing assessable content is the part that does not compress.

Are the adapted versions reliable?

As drafts, yes, with the known weak spot: simplification sometimes strips assessable technical vocabulary. The review pass with the assessment criterion in front of you is mandatory.

Does it work for students learning the language?

It is one of the uses where it shows most: versions with simple syntax and a glossary, keeping the year’s content. Designing the integration itinerary remains the teaching and counseling team’s call.

How do I avoid making it obvious who has which version?

Same visual format for every version and discreet distribution, digital where possible. Good differentiation shows in the whole group working, not in who holds which paper.